Semcon, a product development company specializing in how humans actually use things, conducted a study on people's attitudes toward self-driving cars. People in Sweden, the United Kingdom, the United States, and Germany responded, and nearly half said they had very little trust in autonomous vehicles. Here's a clue why there might be so little trust: 80% of all respondents said that, as pedestrians, they seek eye contact with the driver of a car at an intersection before they cross. Self-driving cars don't have eyes to contact or a nod of recognition to give. How do pedestrians know the sensors see them?

Sure, dolphins use sonar, whiz through the ocean at incredible speeds, and battle sharks. Last week, a study published in Russia's St. Petersburg Polytechnical University Journal: Physics and Mathematics claimed to have recorded two dolphins doing just that. Two Black Sea bottlenose dolphins, named Yasha and Yana, exchanged a series of vocal pulses that resembled "a conversation between two people," wrote the study's author, Vyacheslav Ryabov, a senior researcher at the T. I. Vyazemsky Karadag Scientific Station. What's more, Yana and Yasha were exceedingly polite, listening to one another at turns without interrupting. "As this language exhibits all the design features present in the human spoken language, this indicates a high level of intelligence and consciousness in dolphins, and their language can be ostensibly considered a highly developed spoken language, akin to the human language," Ryabov wrote.

DeepMind, founded in the UK in 2010, created the first computer program to ever beat a professional at the game of Go (AlphaGo), created a DeepRL system to play Atari games at beyond human level performance (DQN), and is engaged in various research projects with the NHS to apply machine learning to radiotherapy planning for head and neck cancers and identification of conditions like age related macular degeneration in optical coherence tomography scans.

How will artificial intelligence systems change the way we live? This is a tough question: on one hand, AI tools are producing compelling advances in complex tasks, with dramatic improvements in energy consumption, audio processing, and leukemia detection. There is extraordinary potential to do much more in the future. On the other hand, AI systems are already making problematic judgements that are producing significant social, cultural, and economic impacts in people's everyday lives. AI and decision-support systems are embedded in a wide array of social institutions, from influencing who is released from jail to shaping the news we see.

Proposals to introduce new legislation which would pardon gay men convicted under historical gross indecency laws will be brought forward "in due course", the government has said. The so-called Alan Turing law could see thousands of men pardoned for crimes of which they would be innocent today. World War Two code-breaker Turing was pardoned in 2013, decades after he was convicted of gross indecency in 1952. A government spokesman said it was "committed" to the proposal. "This government is committed to introducing posthumous pardons for people with certain historical sexual offence convictions who would be innocent of any crime now," the spokesman said.

Famed for its coffee and doughnuts, Tim Hortons is an institution in Canada. Now the restaurant chain is expanding to the UK - but British consumers, it seems, have a lot to learn. On the streets of London, the terms "double double" and "Timbit" baffled most people the BBC spoke to. So we asked Canadians to explain the appeal.

Microsoft has announced a quartet of new initiatives focusing on using artificial intelligence in health care. The company says its researchers are effectively working to "solve" cancer, deploying machine learning techniques for tasks like analyzing tumors and designing new medication regimes. Another projects wants to construct detailed simulations of how cancer develops in different patients' bodies, while one particularly ambitious project -- which Microsoft is calling its "moonshot" effort -- aims to create biological cells that are programmable like computers. Now, creating cells designed to fight cancer is obviously an ambitious task (if it's even possible) and Microsoft is not offering much detail about this particular project. However, personalizing medicine using AI is a much more attainable -- and hopefully effective -- goal.

Fragmentation methods such as the many-body expansion (MBE) are a common strategy to model large systems by partitioning energies into a hierarchy of decreasingly significant contributions. The number of fragments required for chemical accuracy is still prohibitively expensive for ab-initio MBE to compete with force field approximations for applications beyond single-point energies. Alongside the MBE, empirical models of ab-initio potential energy surfaces have improved, especially non-linear models based on neural networks (NN) which can reproduce ab-initio potential energy surfaces rapidly and accurately. Although they are fast, NNs suffer from their own curse of dimensionality; they must be trained on a representative sample of chemical space. In this paper we examine the synergy of the MBE and NN's, and explore their complementarity. The MBE offers a systematic way to treat systems of arbitrary size and intelligently sample chemical space. NN's reduce, by a factor in excess of $10^6$ the computational overhead of the MBE and reproduce the accuracy of ab-initio calculations without specialized force fields. We show they are remarkably general, providing comparable accuracy with drastically different chemical embeddings. To assess this we test a new chemical embedding which can be inverted to predict molecules with desired properties.

We propose an extension to Hawkes processes by treating the levels of self-excitation as a stochastic differential equation. Our new point process allows better approximation in application domains where events and intensities accelerate each other with correlated levels of contagion. We generalize a recent algorithm for simulating draws from Hawkes processes whose levels of excitation are stochastic processes, and propose a hybrid Markov chain Monte Carlo approach for model fitting. Our sampling procedure scales linearly with the number of required events and does not require stationarity of the point process. A modular inference procedure consisting of a combination between Gibbs and Metropolis Hastings steps is put forward. We recover expectation maximization as a special case. Our general approach is illustrated for contagion following geometric Brownian motion and exponential Langevin dynamics.

Determinantal point processes (DPPs) are an important concept in random matrix theory and combinatorics. They have also recently attracted interest in the study of numerical methods for machine learning, as they offer an elegant "missing link" between independent Monte Carlo sampling and deterministic evaluation on regular grids, applicable to a general set of spaces. This is helpful whenever an algorithm *explores* to reduce uncertainty, such as in active learning, Bayesian optimization, reinforcement learning, and marginalization in graphical models. To draw samples from a DPP in practice, existing literature focuses on approximate schemes of low cost, or comparably inefficient exact algorithms like rejection sampling. We point out that, for many settings of relevance to machine learning, it is also possible to draw *exact* samples from DPPs on continuous domains. We start from an intuitive example on the real line, which is then generalized to multivariate real vector spaces. We also compare to previously studied approximations, showing that exact sampling, despite higher cost, can be preferable where precision is needed.